{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/natural-language-generation-by-hierarchical","title":"Natural Language Generation by Hierarchical Decoding with Linguistic Patterns","arxiv_id":"1808.02747","date":"2018-08-08","proceeding":"NAACL 2018 6","authors":["Shang-Yu Su","Kai-Ling Lo","Yi-Ting Yeh","Yun-Nung Chen"],"abstract":"Natural language generation (NLG) is a critical component in spoken dialogue\nsystems. Classic NLG can be divided into two phases: (1) sentence planning:\ndeciding on the overall sentence structure, (2) surface realization:\ndetermining specific word forms and flattening the sentence structure into a\nstring. Many simple NLG models are based on recurrent neural networks (RNN) and\nsequence-to-sequence (seq2seq) model, which basically contains an\nencoder-decoder structure; these NLG models generate sentences from scratch by\njointly optimizing sentence planning and surface realization using a simple\ncross entropy loss training criterion. However, the simple encoder-decoder\narchitecture usually suffers from generating complex and long sentences,\nbecause the decoder has to learn all grammar and diction knowledge. This paper\nintroduces a hierarchical decoding NLG model based on linguistic patterns in\ndifferent levels, and shows that the proposed method outperforms the\ntraditional one with a smaller model size. Furthermore, the design of the\nhierarchical decoding is flexible and easily-extensible in various NLG systems.","url_abs":"http://arxiv.org/abs/1808.02747v2","url_pdf":"http://arxiv.org/pdf/1808.02747v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"natural-language-generation-by-hierarchical","repo_url":"https://github.com/MiuLab/HNLG","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"spoken-dialogue-systems","task_name":"Spoken Dialogue Systems"},{"task_slug":"text-generation","task_name":"Text Generation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1808.02747","atlas_url":"https://app.syntology.ai/?focus=1808.02747","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}